Method for monitoring health state of numerical control machine tool in machining process

Abstract

The invention relates to the technical field of monitoring of a control system in a digital machining process, in particular to a numerical control machine tool machining process health state monitoring method, which comprises the following steps of: determining a machining parameter combination in a health state in a numerical control machine tool machining process, and performing a plurality of groups of machining experiments; collecting various physical quantity signals generated in the experiment process, extracting signal features of the physical quantity signals, constructing a processing process health state monitoring model, training the model, and inputting a processing parameter combination in the actual processing process into the model to obtain model output which is used as a predicted feature value; the difference degree between the characteristic value of the monitoring data and the predicted characteristic value is calculated to serve as the health deviation degree of the machining process, and the health state of the numerical control machine tool in the machining process is determined. The method is suitable for health state monitoring of small-batch processed parts.

Description

Technical Field

[0001] The present invention relates to the technical field of monitoring control systems in digital machining processes, and in particular to a method for monitoring the health status of a machining process of a numerically controlled machine tool. Background Art

[0002] High-value parts, such as aviation parts, have high production and manufacturing costs, and have high requirements for part processing quality. Once an abnormal situation occurs during the processing, if it is not discovered in time, it will lead to rework of parts or even scrapping of parts. More seriously, it may affect the overall progress of the project. In order to avoid cost increases, time losses, etc., and to ensure high-quality processing of high-value parts, it is necessary to monitor the health status of the CNC machine tool processing process in order to detect existing problems in a timely manner and take countermeasures. Existing CNC machine tool processing process monitoring methods use the characteristics of the physical quantities of the processing process and pre-set potential abnormal state types to train artificial intelligence models, or compare the actual processing process data with the standard data under normal conditions, and judge the state of the processing process based on the difference between the two.

[0003] For example, CN116520769B discloses a monitoring method, device, terminal equipment and computer storage medium for a CNC machine tool, which is applied to a twin module in a monitoring system of a CNC machine tool, wherein the monitoring system of the CNC machine tool includes a plurality of monitoring components and a main control module, wherein the plurality of monitoring components, the main control module and the twin module are connected in sequence, and the plurality of monitoring components are connected to the CNC machine tool. Based on the response of the main control module to the target component, the operating status data of the CNC machine tool output by the target component is determined, wherein the target component refers to at least one monitoring component among the plurality of monitoring components; the predicted operating status of the CNC machine tool is determined based on the operating status data, and the actual operating status of the CNC machine tool is monitored based on the predicted operating status.

[0004] CN117784710B discloses a remote status monitoring system and method for CNC machine tools, which monitors and collects machine tool vibration signals in real time through vibration sensors deployed on CNC machine tools, and introduces signal processing and analysis algorithms based on artificial intelligence and deep learning at the back end to analyze the machine tool vibration signals, thereby remotely monitoring the working status of the machine tool in real time, and taking corresponding measures in time to perform maintenance and adjustments based on abnormal working status of the machine tool.

[0005] However, there are at least the following problems in the above-mentioned prior art: the existing CNC machine tool processing process monitoring methods generally focus on evaluating the status of one or several components such as machine tools and cutting tools. Although its evaluation of the monitored object is relatively detailed and accurate, and can give quantitative or semi-quantitative conclusions, it focuses on local or specific abnormal conditions and may ignore other abnormal conditions, and often requires the pre-setting of potential abnormal state types or requires more processing process data under abnormal conditions to train the monitoring model. For application scenarios for high-value parts and high-quality processing, once a problem occurs, no matter where the problem comes from, it may cause serious consequences. Therefore, people hope to first monitor the overall health of the processing process and promptly determine whether there is an abnormality.

[0006] In addition, some other monitoring methods first record a set of data generated under normal processing conditions, and use this as a standard to compare the subsequent monitoring data with it. When the degree of deviation exceeds the set limit, it is considered that the processing state is abnormal. Although this method considers the processing and manufacturing system as a whole rather than focusing on a certain component, it is generally applicable to the mass production of the same type of parts. When processing other types of parts, the standard data needs to be re-collected, so it is not convenient to apply to small-batch production of aviation parts. Summary of the invention

[0007] In order to solve the above problems in the prior art, the present invention proposes a method for monitoring the health status of a CNC machine tool machining process, so as to evaluate whether the machining process is healthy from the perspective of the entire machining process, improve the timeliness of abnormal state warning, and at the same time, there is no need to preset potential abnormal state types in the model training stage. Only a small amount of machining process data under abnormal conditions is required, which can ensure high-value parts to obtain high-quality CNC machining.

[0008] The complete technical solution of the present invention includes: A method for monitoring the health status of a CNC machine tool machining process comprises the following steps: Step (1): Determine the combination of machining parameters that are in a healthy state during the CNC machine tool machining process PP ; Step (2): Combine according to processing parameters PP Conduct multiple processing experiments and collect and record various physical quantity signals generated during the experiments MD ; Step (3): Extract the characteristics of different physical quantity signals from the collected processing physical quantity signals MDC ; Step (4): Construct a healthy state monitoring model for the machining process, using the machining parameter combination when the machining process is in a healthy state PP As the model input, the characteristics of various physical quantity signals of the machining process MDCAs a model output, the model is trained to establish a mapping relationship between the machining parameters and the characteristics of the physical quantity signals of the machining process; Step (5): Combine the processing parameters in the actual processing process PP act Input the model and get the model's output , as the characteristic value of the physical quantity of the machining process that should be monitored when the machining process is in a healthy state, that is, the predicted characteristic value; Step (6): Based on the Mahalanobis distance, the difference between the characteristic value of the monitoring data and the predicted characteristic value is calculated as the healthy deviation of the processing process. DH , the health status of CNC machine tool machining process is determined based on the health deviation.

[0009] Furthermore, in step (1), the processing parameters include the spindle speed of the CNC machine tool n , Feed speed v , cutting depth a p , cutting width a e .

[0010] Furthermore, in step (2), the physical quantity signal includes the current of each axis of the machine tool I , cutting force F ,vibration V ,sound S ,temperature T .

[0011] Furthermore, in step (3), the collected physical quantity signals of the processing process are subjected to denoising and dimensionality reduction processing, and then the features of different physical quantity signals are extracted. MDC .

[0012] Furthermore, in step (3), the characteristics of the physical quantity signal MDC Including time domain features, frequency domain features or time-frequency domain features.

[0013] Furthermore, in step (3), for the current signal, the extracted features include: mean value (MV), root mean square (RMS), standard deviation (SD), peak value (PV), center of gravity frequency (CF) and wavelet packet energy (WPE).

[0014] Furthermore, in step (3), for the cutting force signal, the extracted features include: mean value (MV), root mean square (RMS), standard deviation (SD), peak value (PV), center of gravity frequency (CF) and wavelet packet energy (WPE). FC It consists of the characteristics of cutting force signals in three directions.

[0015] Furthermore, in step (3), for the vibration signal, the extracted features include: root mean square (RMS), standard deviation (SD), peak value (PV), center of gravity frequency (CF) and wavelet packet energy (WPE). VC It is composed of the characteristics of vibration signals in three directions.

[0016] Furthermore, in step (3), for the sound signal, the extracted features include: root mean square (RMS), standard deviation (SD), peak value (PV), center of gravity frequency (CF) and wavelet packet energy (WPE).

[0017] Furthermore, in step (3), for the temperature signal, the extracted features include: peak value (PV).

[0018] Furthermore, in step (6), the healthy deviation of the processing DH for:

[0019] in, Represents the covariance matrix of the dataset.

[0020] Furthermore, in step (6), based on the health deviation, the health status of the CNC machine tool processing process is divided into healthy, sub-healthy, and abnormal states.

[0021] Furthermore, a method for digital machine tool processing is provided, in which a workpiece is processed by a numerically controlled machine tool, and the processing is monitored by the above-mentioned monitoring method.

[0022] Furthermore, the workpiece is an engine blade.

[0023] The advantages of the present invention over the prior art are: The present invention realizes the monitoring of the health status of the machining process of CNC machine tools, can evaluate whether its status is healthy from the perspective of the entire machining process, and improves the timeliness of abnormal state warning; at the same time, there is no need to preset potential abnormal state types in the model training stage, and the model is mainly trained using the machining process data under normal conditions, and only a small amount of machining process data under abnormal conditions is required to assist in determining the healthy deviation range, which solves the problem of unbalanced model training samples caused by the fact that normal data far outnumbers abnormal data in actual production and manufacturing processes; since there is no need to frequently collect standard data, it is suitable for health status monitoring of parts processed in small batches. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 The present invention is a flow chart of the method for monitoring the health status of a CNC machine tool machining process.

[0025] Figure 2It is a schematic diagram of the health status monitoring model of the CNC machine tool machining process of the present invention. DETAILED DESCRIPTION

[0026] The present invention is described in detail below in conjunction with the embodiments and drawings, but it should be understood that the embodiments and drawings are only used to exemplify the present invention and do not constitute any limitation on the protection scope of the present invention. All reasonable changes and combinations within the scope of the inventive concept of the present invention fall within the protection scope of the present invention.

[0027] like Figure 1 As shown, the present invention discloses various steps of a method for monitoring the health status of a CNC machine tool machining process. In this embodiment, the above steps are further described in conjunction with a typical engine blade in an aviation part, including: Step S01, by analyzing the performance of the processing system, determining the processing parameter combination when the processing is in a healthy state, and making a processing parameter table accordingly.

[0028] The performance of the machining system refers to the chatter performance of the machine tool, the cutting mechanical properties of the tool, etc. By analyzing them, we can determine the machining parameter combination that will not cause unhealthy or abnormal machining conditions such as chatter and excessive cutting force. PP , including spindle speed n , Feed speed v , cutting depth a p , cutting width a e etc., that is:

[0029] PP The specific processing parameters contained in can be increased or decreased according to the actual processing conditions, such as the performance of the machine tool and the tool, the material of the workpiece, the process adopted, etc.

[0030] If the machining system is normal, the machining process under these machining parameter combinations is in a healthy state. Since each parameter can have different values ​​under the premise of ensuring that the machining process is in a healthy state, the machining parameter combination PP There are many groups. Assume that the selected processing parameters are p There are 2 types of processing parameters: N group, then the processing parameter table when the processing process is in a healthy state made by analyzing the performance of the processing system is actually a table of size p × N The matrix of .

[0031] Specifically for engine blade manufacturing, the machine tool needs to perform five-axis machining. Therefore, the spindle speed is selected n , Feed speedv , cutting depth a p and cutting width a e As a combination of processing parameters, namely:

[0032] Step S02, performing N groups of processing experiments according to the processing parameter table, and recording various physical quantity signals generated during the experiments through a data acquisition device.

[0033] The processing data acquisition equipment includes various external sensors and acquisition boards and other hardware, such as cutting force, vibration, sound, temperature and other sensors. At the same time, the current CNC system also supports the collection of internal machine tool data such as axis current, spindle power, etc., and provides a data interface for users to obtain relevant data. N Conduct processing experiments and record various physical quantity signals generated during the experiment MD , including the current of each axis of the machine tool I , cutting force F ,vibration V ,sound S ,temperature T etc., that is:

[0034] The physical quantity signals of the experimental process recorded by the data acquisition equipment are time series data. Since a physical quantity signal of a processing process may contain multiple sets of time series data, such as current I It may include current data of multiple axes such as the machine tool x, y, and z axes, cutting force F It may contain cutting force data along multiple directions such as x, y, and z axes, and the same physical quantity may be measured by multiple sensors at the same time. Each sensor will record data within the monitoring time range. Therefore, each physical quantity signal collected is expressed in matrix form, as shown in the following form: , wait.

[0035] in, r , s etc. are the number of time series contained in each physical quantity signal; T I , T F etc. are the lengths of the corresponding time series data of the physical quantity. These processing physical quantity signals reflect the specific state of the processing.

[0036] In this embodiment, specifically for engine blade manufacturing, physical quantities of the processing process are selected for monitoring: current, cutting force, vibration, sound and temperature.

[0037] Step S03, extracting the characteristics of physical quantity signals of different machining processes respectively through digital signal processing and other technologies.

[0038] The collected physical quantity signals of the processing process are processed by denoising and dimensionality reduction. Then, according to the amplitude, frequency and other characteristics of different signals in the processing process, the time domain, frequency domain or time-frequency domain characteristics of different physical quantity signals are extracted through digital signal processing and other technologies, which are recorded as IC = ( IC 1, IC 2, ..., IC m ), FC = ( FC 1, FC 2, ..., FC n ), etc., so all the features extracted from the monitoring signal are recorded as:

[0039] During the manufacturing process of engine blades, the monitored physical quantities are denoised and the following features are extracted: For the current signal, the extracted features include: mean value (MV), root mean square (RMS), standard deviation (SD), peak value (PV), center of gravity frequency (CF) and wavelet packet energy (WPE). Since the current signal includes the spindle current signal and the current signals of each feed axis, the features of the current signal IC It is composed of the characteristics of the spindle current signal and the characteristics of each feed axis current signal, which are expressed as follows: Table 1 Characteristics of current signal

[0040] The sign of the current signal characteristic IC The subscripts of (and the elements it contains) are explained as follows: 0 represents the machine tool spindle, f Represents the feed axis number (for example f = 1 for the first feed axis). Therefore, the characteristic of the current signal is IC = ( IC 0, IC 1, ..., IC f , ..., IC 5).

[0041] For the cutting force signal, the extracted features include: mean value (MV), root mean square (RMS), standard deviation (SD), peak value (PV), center of gravity frequency (CF) and wavelet packet energy (WPE). Since the cutting force signal includes signals along the x, y and z directions, the features of the cutting force signal are FC It consists of the characteristics of cutting force signals in three directions, which are expressed as follows: Table 2 Characteristics of cutting force signal

[0042] Sign of cutting force signal characteristics FC The subscripts of (and its elements) are explained as follows: x, y, and z represent the features extracted from the cutting force signal along the x, y, and z directions, respectively. Therefore, the features of the cutting force signal FC = ( FC x , FC y , FC z ).

[0043] For vibration signals, the extracted features include: root mean square (RMS), standard deviation (SD), peak value (PV), center of gravity frequency (CF) and wavelet packet energy (WPE). Since the vibration signal includes signals along the x, y and z directions, the features of the vibration signal VC It consists of the characteristics of vibration signals in three directions, which are expressed as follows: Table 3 Characteristics of vibration signals

[0044] Sign of vibration signal characteristics VC The subscripts of (and its elements) are explained as follows: x, y, and z represent the features extracted from the vibration signal along the x, y, and z directions, respectively. Therefore, the features of the vibration signal VC = ( VC x , VC y , VC z ).

[0045] For sound signals, the extracted features include: root mean square (RMS), standard deviation (SD), peak value (PV), center of gravity frequency (CF) and wavelet packet energy (WPE), that is, SC = ( RMS S , SD S , PV S , CF S , WPE S1 , WPE S2 , WPE S3 ). Among them, the subscript S The object representing the extracted features is the sound signal.

[0046] For temperature signals, the extracted features include: peak value (PV), i.e. TC = ( PV T ). Among them, the subscript T The object representing the extracted features is the temperature signal.

[0047] Combine the features of various physical quantity signals to obtain all the features extracted from the monitoring signal:

[0048] Step S04: construct a machining process health status monitoring model based on a neural network artificial intelligence algorithm.

[0049] like Figure 2 As shown. The processing parameters combination when the processing is in a healthy state PP As the model input, the characteristics of various physical quantity signals of the machining process MDC As the model output, the model is trained to establish the mapping relationship between the processing parameters and the characteristics of the physical quantity signal of the processing process. N The number of samples in the training set is N The established mapping relationship is used to characterize the processing system in a healthy state.

[0050] Step S05, inputting the processing parameter combination used in actual part manufacturing into the model, and calculating the degree of difference between the physical quantity characteristics predicted by the model and the characteristics of the monitored physical quantity.

[0051] Combine the processing parameters in the actual processing process PP act Input the trained model and get the output of the model .

[0052] For aviation blade processing, its output When the machining process is in a healthy state, the machining parameters are combined PP actThe characteristic values ​​of the physical quantities of the machining process that should be monitored during machining are the predicted characteristic values. Since the model represents the machining system in a normal state, if the actual machining process is normal, the characteristic values ​​of the monitoring data should be similar to the predicted characteristic values. Otherwise, it means that there may be an abnormality in the machining process. Data similarity description algorithms such as Mahalanobis distance can derive the degree of difference between the two and use it as the healthy deviation of the machining process. DH , the following is calculated using the Mahalanobis distance as an example DH :

[0053] in, Represents the covariance matrix of the dataset.

[0054] For aviation blade processing, the healthy deviation is:

[0055] And according to DH e The following table shows DH The value range determines the state of the machining process, where DH The values ​​are normalized to [0, 1).

[0056] Step S06, providing a graded evaluation result of whether the processing process is in a healthy, sub-healthy, abnormal, etc. state according to the degree of difference of the prediction / monitoring characteristics.

[0057] Healthy Deviation DH The degree to which the actual processing process deviates from the healthy state is quantified. DH The larger the value, the greater the degree of deviation of the processing process from the healthy state. Based on this, the corresponding values ​​of the processing process in the healthy, sub-healthy, and abnormal states are determined. DH The value range is shown in Table 1: Table 4 CNC machine tool processing health status classification evaluation standard

[0058] in, DH a < DH b < DH c < DH d According to the actual processing DHThe state corresponding to the value interval gives the graded evaluation results of the processing process in healthy, sub-healthy, abnormal and other states. According to the evaluation results, production and manufacturing personnel can respond to potential problems in the processing process in a timely manner. For example, "sub-health" means that the processing system needs to be checked, but not necessarily immediately, and "abnormal" means that the processing needs to be stopped immediately for inspection. The graded evaluation standards and corresponding personnel behavior codes can be appropriately adjusted according to the specific production scenarios to formulate production operation specifications that meet the actual factory needs. At the same time, it is also possible to write programs to realize functions such as automatic alarm of the processing system and automatic shutdown in serious abnormal conditions to avoid major losses.

[0059] Table 5 Health status classification evaluation criteria for engine blade CNC machine tool machining process

[0060] For example, for aviation blade processing, DH e = 0.10, the machining process is healthy; DH e = 0.36, the processing is sub-healthy; DH e = 0.78, the machining process is abnormal.

[0061] Depending on the specific manufacturing process and the sensors used, technicians can adjust the type and quantity of extracted monitoring signal features and the grading evaluation criteria for the health status of the processing process to suit specific scenario requirements.

[0062] Through the method described in this embodiment, the health status of the CNC machine tool processing process is monitored, and its health can be evaluated from the perspective of the entire processing process, thereby improving the timeliness of abnormal state warning. There is no need to preset potential abnormal state types during the model training stage, and the model is mainly trained using the processing data under normal conditions. Only a small amount of processing data under abnormal conditions is required to assist in determining the range of healthy deviations, which solves the problem of imbalanced model training samples caused by the fact that normal data far outnumbers abnormal data in the actual production process. Since there is no need to frequently collect standard data, it is suitable for health status monitoring of parts processed in small batches. At the same time, the proposed method gives the evaluation results in the quantitative form of the healthy deviation of the processing process, which provides a basis for developing a monitoring program based on the evaluation results, helps to improve the automation and intelligence level of the production site, saves production costs in the long run, and improves production efficiency.

[0063] The above applications are only some implementation methods of the present application. For those skilled in the art, several modifications and improvements can be made without departing from the creative concept of the present application, and these all belong to the protection scope of the present application.

Claims

1. A method for monitoring the health status of a CNC machine tool machining process, characterized in that: The steps include: Step (1): Determine the combination of machining parameters that are in a healthy state during the CNC machine tool machining process PP ; Step (2): Combine according to processing parameters PP Conduct multiple processing experiments and collect and record various physical quantity signals generated during the experiments MD ; Step (3): Extract the characteristics of different physical quantity signals from the collected processing physical quantity signals MDC ; Step (4): Construct a health status monitoring model for the machining process, using the machining parameter combination when the machining process is in a healthy state PP As the model input, the characteristics of various physical quantity signals of the machining process MDC As a model output, the model is trained to establish a mapping relationship between the machining parameters and the characteristics of the physical quantity signals of the machining process; Step (5): Combine the processing parameters in the actual processing process PP act Input the model and get the model's output , as the characteristic value of the physical quantity of the machining process that should be monitored when the machining process is in a healthy state, that is, the predicted characteristic value; Step (6): Based on the Mahalanobis distance, the difference between the characteristic value of the monitoring data and the predicted characteristic value is calculated as the healthy deviation of the processing process. DH , the health status of CNC machine tool machining process is determined based on the health deviation.

2. A method for monitoring the health status of a CNC machine tool machining process according to claim 1, characterized in that: In step (1), the processing parameters include the spindle speed of the CNC machine tool n , Feed speed v , cutting depth a p , cutting width a e .

3. A method for monitoring the health status of a CNC machine tool machining process according to claim 2, characterized in that: In step (2), the physical quantity signal includes the current of each axis of the machine tool I , cutting force F ,vibration V ,sound S ,temperature T .

4. A method for monitoring the health status of a CNC machine tool machining process according to claim 3, characterized in that: In step (3), the collected physical quantity signals of the processing process are subjected to denoising and dimensionality reduction processing, and then the features of different physical quantity signals are extracted. MDC .

5. A method for monitoring the health status of a CNC machine tool machining process according to claim 4, characterized in that: In step (3), the characteristics of the physical quantity signal MDC Including time domain features, frequency domain features or time-frequency domain features.

6. A method for monitoring the health status of a CNC machine tool machining process according to claim 5, characterized in that: In step (6), the healthy deviation of the processing process DH for: in, Represents the covariance matrix of the dataset.

7. A method for monitoring the health status of a CNC machine tool machining process according to claim 6, characterized in that: In step (6), based on the health deviation, the health status of the CNC machine tool processing process is divided into healthy, sub-healthy, and abnormal states.

8. A method for data machine tool processing, characterized in that: When a workpiece is processed by a numerically controlled machine tool, the processing is monitored by using the monitoring method described in any one of claims 1 to 7.

9. The method according to claim 8, characterized in that The workpiece is an engine blade.

Images